Submitted to The Astrophysical Journal
Improving the Resolution of X-Ray Telescopes with Occulting
Craig J. Copi^ and Glenn D. Starkman^
email@example.com & firstname.lastname@example.org
Improving angular resolution is one of X-ray astronomy's big challenges.
While X-ray interferometry should eventually vastly improve broad-band angular
resolution, in the near-term, X-ray telescopes will sacrifice angular resolution for
increased collecting area and energy resolution. Natural occultations have long
been used to study sources on small angular scales, but are limited by short tran-
sit times and the rarity of transits. We describe here how one can make use of an
X-ray Occulting Steerable Satellite (XOSS) to achieve very-high resolution of X-
ray sources, conventional X-ray telescopes. Similar occulting satellites could also
be deployed in conjunction with future space observatories in other wavebands.
Subject headings: space vehicles — occultations — X-rays: general
One challenge of X-ray astronomy is the relatively low photon fiuxes from target sources.
X-ray telescope builders are forced to trade angular resolution against increased collecting
area. Thus, while the diffraction limit of a 1.2 m aperture X-ray telescope (such as Chandra)
is 0.3 mas at 1 keV, the 0.5 as reality is worse than what is routinely achieved at longer
wavelengths. While development proceeds on space-based X-ray interferometers, near-future
X-ray missions such as Constellation X plan to increase effective area at the price of reduced
angular resolution. MAXIM Pathfinder, which will afford 100 /xas resolution , hopes to
launch between 2008 and 2013; however, it will have an effective area of only 100 cm^.
^Department of Physics, Case Western Reserve University, Cleveland, OH 44106-7079
- 2 -
In the interim, angular resolution can be substantially improved for bright, relatively
stable sources. The general technique to do this is well-known — eclipse mapping. When a
body such as the moon transits a telescope's field-of-view (FOV), it occults different sources
within the FOV at different times. By carefully measuring the photon count rate as a function
of time during the transit, one can reconstruct the projection of the surface brightness in
the FOV onto the path of the occulter.
Deployment of large occulting satellites has been discussed for the optical and near infra-
red (Adams et al. 1988; Schneider 1995; Copi & Starkman 1998; Copi & Starkman 2000),
for both planet finding and high-resolution astronomy. However, occulting satellites are
particularly well-suited for use in the X-ray. In the X-ray, the diffraction of the photons
by the sateUite can generally be neglected (see section 3 below). The optimal size and
placement of the sateUite are therefore governed chiefly by one's ability to accurately position
the satellite and station-keep with respect to the line-of-sight to the source. The achievable
resolution is then determined primarily by the telescope collecting area and by the level to
which one can zero the XOSS-telescope relative velocity.
We consider an X-ray telescope at the the Earth-Sun system's second Lagrange point
(L2) (such as Constellation X). We discuss the size and steering of a XOSS in section 2. In
section 3 we present the two-source angular resolution as a function of the XOS'S'-tclcscope
relative angular velocity and of the photon count rate. In section 4 we reconstruct some test
sources. Specific classes of sources are discussed briefiy in section 5. Section 6 contains our
2. Building and Using a XOSS
Locating XOSS: All current and some future X-ray telescopes use Earth orbit. Because
orbital velocities are so high there, it is more difficult to use the approach we describe here;
we will not discuss these further. Other X-ray telescopes, such as Constellation X, will
be located at L2 or other low-acceleration environment (e.g. Earth-trailing and drift away
orbits). We focus on these.
We have previously discussed in detail placing a large occulter at L2 (Copi & Starkman 2000).
Here we highlight the main points. Orbits around L2, both in and out of the ecliptic, have
periods r ~ 6 months, nearly independent of their size (for d < 10"^ km). Consider an X-ray
telescope at L2 and a XOSS orbiting it at a distance d. The typical orbital acceleration
is a = An'^d/r'^. Suppose initially XOSS and the telescope are at relative rest. We require
their relative velocity through a transit to be less than some v, the smaller v, the better the
expected angular resolution. For a satellite of size L, the transit duration is t — L/v, so we
need at < v. In terms of the angular velocity, /i — v/d, the constraint is
VaL If L \ /lO^ m\ mas
;,>_ = 0.26^(— j(— j— . (1)
This is the typical relative angular velocity assuming no orbital adjustments are made.
Much smaller relative angular velocities can be achieved. Drifts of 1 m off the hne-of-site
can take days, if the relative-velocity perpendicular to the line-of-sight is initially zeroed
Building XOSS The attenuation length of X-rays in matter is well known (Hagiwara et al. 2002).
Except in hydrogen, it is approximately 3 x 10~^ g/cm^ at 1 keV, 10~^ g/cm^ at 10 keV,
and 0.2 g/cw? at 100 keV. At a density of 3 g/cm^, these represent thicknesses of 1, 30 and
600 microns respectively. An occultcr needs to be ~ 3 attenuation lengths thick, and so for
a 10 m square ~ 1 kg to operate at 1 keV, 30 kg at 10 kcV. Even at 100 keV a 3 m x 3 m
lead film three attenuation lengths thick would be only 60 kg.
The ideal size for XOSS depends on the telescope aperture, the accuracy with which one
can position the occulter, and the telescope field-of-view (FOV) . A typical X-ray aperture,
which is the minimum occulter size, is ~ 1 m. Next we estimate how well we can find
XOSS's position relative to the line-of-site. Let r be the telescope- XOS'S' separation. With
a diffraction-limited optical telescope of diameter d on one spacecraft one can establish the
relative position of the other to
-. ^ A ^ ^ r A/400 nm
5x < 1.2r- = 0.5 m— — '— 2
d 1000 km d/\m ^ '
1 m accuracy at 1000 km separation therefore requires a 50 cm finder scope. With isotropic
scattering from the target, and a reflecting area of 1 m^, sunlight results in a flux of 3 x
10^/m^/s, at 1000 km, suggesting that sufficient light is available to determine relative
positions to ~ 1 m or less.
We must also be able to reduce the spacecrafts' relative velocity to <C 1 m/s. In
principle, two position determinations each with error of Ax, made a time t apart, determine
the velocity within Av ~ -\/2Ax/i (assuming no error in t). The ability to reduce |Ai>| is
then constrained by the precision control on the impulse rockets, and by the maximum time
between position determinations. This time is limited by the orbital accelerations, but is
sufficiently long (> 10^ s) at L2.
Existing or planned X-ray telescopes that could be used with XOSS have large FOV.
If XOSS covers only part of the FOV, the rest will contribute a large background. This
can be mitigated by monitoring each telescope resolution element. For Constellation- X with
~ 15 as this would still require a 70 m XOSS at 1000 km separation, though the full
resolution element need not be occulted since some background is acceptable. Occulters of
this scale have been previously considered (Copi & Starkman 2000). Still, this provides an
important constraint on the size and distance of XOSS; alternatively it suggests that an
X-ray telescope with somewhat higher resolution would be a better detector for XOSS.
Steering XOSS To transit multiple sources it will be necessary to accelerate the satellite
both to retarget, and to velocity-match. For an arcal density of just 1.5 x lO^"' g/cm^
(five attenuation lengths as 1 keV), solar radiation pressure causes an acceleration of only
4 X 10~^ m/ s^. All velocity adjustments will therefore require rockets.
A change Av in the satellite's velocity is related by momentum conservation to the mass
of propellant ejected and the velocity of ejection:
If is the desired number of velocity changes, then we must keep Ampropeiiant < ^sat/A^-
Ion engines currently have t'ejection — 30 km/s, thus
A^ < 30 km/s/A^;saf (4)
For satellites separated by 1000 km near L2, relative velocities are only i>sat = C'(10~^ km/s),
suggesting that propellant supply is not a concern.
The angular resolution of the XOSS-telescope system will come from probing the lightcurve
during a transit. Consider first identifying a binary source. For 0.5 keV X-rays edge diffrac-
tion from XOSS causes oscillations of order 5 cm in the aperture plane of a telescope 1000 km
away. A 1-m aperture telescope averages over many such oscillations leading to very small
deviations from the geometric shadow, which are further reduced at higher energies. Deter-
mining the lightcurve therefore reduces to calculating the area of the telescope not shadowed
by XOSS. An observation consists of a sequence of measurements of the integrated lightcurve
between times and tj. We would like to find the minimum separation of two sources that
can be distinguished from a single source, using such observations. We therefore calculate
the probability of misidentifying a binary source as a single source. This probability depends
on the angular velocity of XOSS as it transits, and on the photon count rate. (Through-
out we quote photon rates in the detector, not at the the telescope aperture. The X-ray
- 5 -
1 10 100 1000 10^ 10^
Signal photon rate (s )
Fig. 1. — The minimum angular separation of two X-ray sources resolvable at the 95% C.L.
Limits are shown for intensity ratios p = 1 (solid), 1/3 (dashed), and 1/10 (dashed-dotted).
The three sets of curves are for ij,± = 10 mas/s (upper), 1 mas/s (middle), and 0.1 mas/s
(lower). The total photon from both sources detected in (not incident on) the unocculted
telescope. A background equal to the flux of the two sources is assumed. Here d is the
telescope's effective area includes both the geometric collecting area and the efficiency of the
X-ray detector.) The results for the 95% confidence limits as a function of the intensity in a
1.2 m aperture telescope for p = 1, 1/3, and 1/10 and for /i_|_/(mas/s) = 10, 1, and 0.1 are
shown in figure 1. In producing figure 1 we assumed a uniform response over the surface of
the telescope. A more complicated response function may improve resolution slightly.
This analysis assumed a single transit by a square occulter, yielding only one projection
of the source positions. One needs multiple projections to resolve a 2-d source. This could
be accomplished by multiple transits, or slits in XOSS tilted to mimic multiple passes in a
4. Image Reconstruction
Several techniques have been applied to image reconstruction from natural eclipses. We
have focused on the eclipse mapping method (Baptista & Stcincr 1991). EMM is a maximum
entropy technique, thus it can oversmooth the image or introduce spurious sources depending
on the weighting between the entropy and the constraints. We have implemented an improved
EMM algorithm (Baptista & Steiner 1993), in which we maximize the quality function
where S is the entropy, the usual definition of is
used, C[x) — [x — a;aim)/2;aim;
measures the correlations in the residuals, and p controls the weighting between the entropy
and the constraints. We anneal in p from 100 to 10"^, to allow first the entropy to smooth
the image then the constraints to sharpen the important features. In practice we tend to
oversmooth, leading to a conservative estimate of the attainable resolution.
We modified the EMM algorithm. In the geometric limit, sources contribute to changes
in the lightcurve while they are occulted. We therefore give more weight to pixels that
are causing the light curve to change. A minimum pixel intensity is chosen; pixels above
threshold count fully in the fit while those below the minimum are averaged over. Again we
anneal from a high value for this intensity threshold to a low value.
To start the reconstruction we could use a low resolution image. For the test cases
considered here we instead bootstrap by initially performing a non-negative least squares fit.
Most of the signal is typically placed in a few very bright pixels — to speed up convergence.
Fig. 2. — The image reconstruction of (a) Gaussian point sources and (d) filaments. Re-
constructed images after 4 passes (b and e) and 16 passes (c and f) show the reconstruction
abihty of XOSS.
we presmooth the image with a Gaussian. The smoothing scale must be chosen with some
care. Smoothing on the scale of a few pixels worked well for our test cases.
Two test cases have been considered (Fig. 2 panels (a) and (d) contain the original
sources). The first contains three Gaussian sources of equal integrated intensity but peak
intensities in a 4:2:1 ratio. The second contains filaments radiating from a bright spot. The
reconstructions are performed for a 10^7/snapshot source. XOSS moves 1 pixel per snapshot.
The square satellite provides information along only 1 direction per transit. Thus we consider
(b,e) 4 and (d,f) 16 passes for each reconstruction. To extract maximal information from N
passes, each pass the occulter travels at an angle 9 = tt/N to the previous pass.
For the Gaussian source the reconstructions are oversmoothed. By 4 passes (b) we
can resolve at least two sources; with 16 passes (c) all three sources can be resolved. For
the filamentary source the 4 passes (e) reconstruct the bright spot well. With 16 passes
(f) the filaments are resolved. Even though the reconstruction technique involves Gaussian
smoothing, non-Gaussian structures can be resolved.
This is an exciting time for X-ray astronomy. The Chandra and XMM telescopes have
been inserted into Earth orbit. Other major observatories are being planned for this decade:
Astro-E2, and XEUS will enter Earth-orbit, while Constellation X will be placed at L2.
MAXIM Pathfinder — the first astronomical X-ray interferometer — may launch between 2008
and 2013 into a drift-away orbit; to be followed eventually by MAXIM. This may seem a
remarkable proliferation, but each mission has its own emphasis. Chandra is the only high
angular resolution telescope, with ~ 0.5 as, and thus has the lowest effective area. The
other missions (excepting the interferometers) opt instead for large effective area.
The luminosity of X-ray sources varies greatly. Black holes in the cores of nearby galaxies
£bh ~ 10=^^"^° erg/s = 6.2 x 10^^"^^ keV/s (7)
in the 0.2-2.4 keV energy range. This leads to a photon rate at the detector of
= 6.5 X 100-2 -g^ev
where E'keV is the X-ray energy in keV, d^pc is the source distance in Mpc and ^looo is the
effective area of the telescope in units of lOOOcm^. An AGN, Seyfert galaxy, or the core of an
X-ray cluster can be much more luminous >Cagn ~ 10^° erg/s; however, since d ^ 100 Mpc
Tagn = 6.5 X io(-2to+2) lO^^keV
Galactic microquasars are less luminous vCmicroquasar ~ 10^^ erg/s, but, since they are closer
rmicroquasar — 6.5 X 10 Z S . (10)
Great improvements in resolution are readily accessible (figure 1). Sub-milliarcsecond
resolution can be obtained for sources with photon rates F > 1000 s~^. For a single Con-
stellation X module, which will have an effective area of about 3, 750 cm^, the brightest
AGNs, X-ray cluster cores, and galactic black holes will have F ^ 200 s~^ so we can obtain
pa 5 mas.
This technique has clearest application to steady sources. The occulter transit time is
f F\ f mas/s\ , ,
^transit = 1000 S ( - J ( ^^^^ 1 , (H)
where F is the telescope FOV or resolution, whichever is less. The required source-stability
timescale is ^transit times the number of passes. Many extragalactic sources are sufficiently
stable. These include galaxy clusters, AGNs, quasars, and many X-ray jets; so too are
supernovae. Galactic sources, however, often vary more quickly. It may nevertheless be
possible to learn something about them using this technique. In particular, optimization
of the aperture (e.g. by use of coded masks) may allow us to significantly reduce both the
required transit time and the required number of transits to obtain a well- resolved 2-d image.
We have found that a XOSS can lead to tremendous improvements in angular resolution.
The current trend of increasing the effective area of X-ray telescopes at the expense of
angular resolution until X-ray interferometry becomes viable, meshes well with the benefits
gained by including a XOSS in the mission. For the high Earth-orbit X-ray telescope a
moderate improvement in angular resolution over an appreciable fraction of the sky can be
achieved through the use of a XOSS, however the orbital mechanics are challenging and the
scientific pay-off may be modest. A technology demonstration mission along these lines may
nevertheless be worthwhile.
XOSS is potentially most interesting for telescopes at L2 and other low-acceleration
environments (e.g. Earth-trailing or drift-away orbits) where 1-10 mas resolution should be
achievable for a wide range of sources. The ideal telescope would have an angular resolution
equal to the angle subtended by the XOSS. Further investigations arc necessary to determine
whether XOSS is well suited to Constellation X, or whether a dedicated telescope with better
angular resolution would be merited. Either way, it seems likely that XOSS could provide
a natural scientific precursor to interferometric missions like MAXIM Pathfinder, increasing
the angular resolution of X-ray telescopes by one-to-two orders-of-magnitude at relatively
modest incremental expense.
The authors thank A. Chmielewski for support, A. Babul and M. Dragovan for many
useful comments and suggestions, N. Choudhuri for helpful suggestions on statistical tests,
P. Gorenstein and W. Zhang for comments on a preliminary version of the manuscript and
C. Covault and S. Rodney for recent input. This work was supported by a DOE grant to
the theoretical particle- astrophysics group at CWRU, and a phase I grant from NIAC.
Adams, D.J. et al., 1988, ApJ, 327, L65
Baptista, R. & Steiner, J.E., 1991, A&A, 249, 284
Baptista, R. & Steiner, J.E., 1993, A&A, 277, 331
Copi, C.J. & Starkman, G.D., 1998, Proceedings of SPIE, 3356, March 25, 1998, Kona, HI
Copi, C.J. & Starkman, CD., 2000, ApJ, 532, 581
Hagiwara, K. et. al., 2002 Phys. Rev. D, 66, 010001.
Schneider, J., 1995, Proceeding of the workshop "Detection and Study of Terrestrial Extra-
solar Planets", May 15-17, Boulder, CO, unpublished
This preprint was prepared with the AAS IM^jX macros v5.0.